Facility layers: Floor · Air · Water · Fire · Power

Floor: “Will it fit?”

What you see: rooms, tile grid, hot and cold aisles, racks with U used, kW against budget and weight per rack. Aisles alternate by row unless your Racks sheet says otherwise (front_faces A or B).

Checks pinned in 3D: U overlaps, overweight racks, kW over the rack budget, unknown models shown as plain boxes to identify later.

From the sheet: Racks (rack, row, position, height_u, width_in, front_faces, max_kw) and Devices (rack, u, model, height_u).

A hot aisle / cold aisle layout tool should take rows and racks and give you aisles, not a drawing exercise. Fill the Racks sheet with row and position; the room, the aisles and the rack budgets appear. Change front_faces and the aisle flips.

Air: “A cooling unit stops. How many minutes?”

What you see: rack inlet and exhaust temperature, a floor heat map, airflow particles; the hottest inlet as a KPI in Monitor.

What-if: fail a CRAH, chiller or AC unit and inlet temperatures climb; minutes to 27, 32 and 35 °C per rack, estimated by a zonal model and labelled as such. Not CFD.

From the sheet: Cooling (name, type, capacity_kw, serves, model), plus temperature history in the Readings sheet if you have it. Heat above cooling capacity is flagged at import.

Water: “Which racks does this leak take out?”

What you see: chilled-water loops, CDUs, manifolds, coolant return temperature, flow particles.

What-if: a CDU stops; a manifold leaks: litres released (estimated), the isolation valve closes, racks are taken offline. Built for small rooms adding a liquid-cooled GPU row next to air-cooled racks: both live in one model.

From the sheet: Cooling rows of liquid types (CDU, loop) with serves listing the racks.

Liquid cooling planning for a small data center starts with one question: which racks go on the loop, and what happens to them when the CDU stops. Model it before the pipes are ordered.

Fire: “What is the sequence, and who is affected?”

What you see: detection zones by state, detectors, suppression cylinders.

What-if: smoke in a zone, then an early-warning alert, alarm, HVAC interlock, agent release and EPO, as a timeline with the racks and services in the zone listed. Fire is a modelled layer here, not just an alarm input: you see the sequence, not only a red icon.

From the sheet: zones are assumed (one zone per room) until you add them; the import report says so.

Power: “Lose feed A at peak. Who goes dark?”

What you see (3D): rack kW columns, PDU strips coloured by load, power flow along the power cords.

What you see (Diagram): an auto-built one-line (utility, generator and ATS, UPS, panel, PDU, device PSU) with A and B paths in colour, load % on every node and UPS runtime: your runtime_min or battery_kwh from the sheet first, then the catalog, otherwise estimated and labelled.

What-if: click any node, press Fail, and see what goes dark, what survives on one feed, which services stop and what overloads. Simulate it replays the failure through the full model: batteries, generator start, breaker curves, thermal transients, tokens lost.

Checks: any node over 80 % of its rating; per-phase load against 80 % and phase imbalance above 15 % (when the sheets give phases); single-corded or same-feed devices; UPS runtime below 5 minutes at current load.

From the sheet: Power (name, type, rating, fed_from, feed, rack, voltage, phases, phase, runtime_min, battery_kwh, model) and Cables of type power, or the power_cords shortcut (A, B or AB) on Devices.

Rack power capacity planning without a tool means a spreadsheet of nameplate watts and a guess at derating. Here the Devices sheet gives per-rack kW from the catalog, the Power sheet gives breaker and PDU load, and the 80 % rule is checked for you. UPS failure impact analysis is one click: fail the UPS, read the list of orphaned devices, stopped services and overloaded PDUs. Try the rack power calculator or read UPS failure impact analysis in 20 minutes.